01OpenAI· This week
OpenAI outlines safety risks and safeguards for long-horizon AI models
OpenAI describes new safety risks tied to long-horizon AI models in a newly published report.
Trust76
HypeLow hype
Archive
01 02 02Moonshot AI· This weekMoonshot AI releases Kimi K3, a 2.8T parameter MoE model with open weights scheduled for July 27
Moonshot AI released Kimi K3, a 2.8 trillion parameter Mixture-of-Experts model slated for open-weights release on July 27, 2026.
Trust78HypeLow hype03 03Quotation· This weekOpenAI considered releasing a GPT-3-level local model in 2022 to deter competitors
OpenAI discussed releasing a GPT-3-level language model capable of running on consumer hardware in 2022.
Trust79HypeLow hype04 04Simon Willison· This weekClaude Code confirmed using Bun’s Rust port in production
Claude Code v2.1.181 and later versions embed a Rust-based port of Bun.
Trust79HypeLow hype05 05Moonshot AI· This weekMoonshot AI releases Kimi K3 open source model, touting frontier-level performance
Moonshot AI released Kimi K3, an open source model it says achieves frontier-level performance on internal evaluations.
Trust71HypeSome hype06 06OpenAI· This weekOpenAI CFO proposes scorecard to measure AI ROI
OpenAI's CFO introduced a scorecard framework to assess AI return on investment.
Trust74HypeLow hype07 07Moonshot AI· This weekMoonshot AI releases Kimi K3 as the largest open-weights model to date with 2.8T parameters and 1M-token context
Moonshot AI launched Kimi K3 as a frontier-class open-weights model with 2.8 trillion parameters and a 1 million-token context window.
Trust78HypeLow hype08 08TechCrunch — AI· This weekMoonshot’s Kimi K3 expected to be China’s largest open-weight AI model with 2T–3T parameters
Moonshot AI’s Kimi K3 is expected to close the performance gap with Anthropic’s Opus 4.8.
Trust69HypeSome hype09 09New model release· This weekThinky releases Inkling, a 975B-parameter Apache 2.0-licensed multimodal MoE model with 1M-token context
Thinky introduced Inkling, its first open-weights foundation model family, with a 975B total/41B active parameter Mixture-of-Experts architecture supporting text, image, and audio inputs and up to 1M-token context.
Trust78HypeLow hype10 10Startup release· This weekThinking Machines Labs releases open-weight AI model Inkling with 975B parameters and native multimodal training
Thinking Machines Labs released Inkling, its first proprietary AI model, as an open-weight system with 975 billion total parameters and a 41 billion active parameter design for efficiency.
Trust79HypeLow hype11 11New open small language modelsBilibili releases Index-1.9B series of open small language models with technical details and benchmarks
Bilibili released Index-1.9B, a series of four open small language models with 1.9B non-embedding parameters each.
Trust84HypeLow hype12 12OpenAI releaseOpenAI releases GPT-5.6 family with Luna, Terra, and Sol variants
OpenAI’s GPT-5.6 family introduces three variants—Luna, Terra, and Sol—with input/output pricing ranging from $1/$6 to $5/$30 per 1M tokens.
Trust78HypeLow hype13 13OpenAIOpenAI releases GPT-5.6 family with three variants and claims cybersecurity and coding leadership
OpenAI unveiled GPT-5.6, a new family of models with three variants: Sol (workhorse), Terra (intermediate), and Luna (budget).
Trust75HypeSome hype14 14Simon Willison — everythingOpenAI clarifies ChatGPT Work cloud and desktop data separation at launch
OpenAI's documentation states cloud Work conversations are not visible in the desktop app at launch.
Trust79HypeLow hype15 15OpenAIOpenAI and Deutsche Telekom announce multi-year partnership to integrate AI across telecom operations
Deutsche Telekom and OpenAI announced a multi-year partnership to integrate AI across Deutsche Telekom's operations.
Trust71HypeSome hype16 16OpenAIOpenAI announces ChatGPT Work, an agentic system for multi-step tasks across apps and files
OpenAI describes a new agentic system called ChatGPT Work designed to persist across multi-hour projects and execute actions across users' apps and files.
Trust71HypeSome hype17 17Research announcementGemma 4 released with multimodal, encoder-free architecture and reasoning traces
Gemma 4 introduces a new generation of open-weight, natively multimodal language models with dense and Mixture-of-Experts architectures ranging from 2.3B to 31B parameters.
Trust79HypeLow hype18 18Model releaseTencent releases Hy3, a 295B-parameter MoE model with 21B active parameters and 256K context
Tencent’s Hy3 is a 295B-parameter Mixture-of-Experts model with 21B active parameters and 3.8B MTP layer parameters.
Trust79HypeLow hype19 19Tool-use regressionNewer Anthropic models increasingly emit malformed tool calls for third-party edit schemas
Newer Anthropic models (Opus 4.8, Sonnet 5) sometimes add invented keys to the edits[] array when calling third-party edit tools, causing schema validation failures.
Trust79HypeLow hype20 20Model releaseClaude Sonnet 5 launches with 1M token context window, new tokenizer, and adaptive thinking enabled by default
Claude Sonnet 5 is positioned by Anthropic as having performance close to Opus 4.8 at lower prices.
Trust79HypeLow hype21 21AnthropicAnthropic releases Claude Sonnet 5 with stronger agentic capabilities and lower pricing
Claude Sonnet 5 is Anthropic’s latest mid-size model, offering stronger agentic capabilities than its predecessor.
Trust79HypeLow hype22 22OpenAIOpenAI engineers fix 18-year-old infrastructure bug using core dump epidemiology
OpenAI engineers diagnosed rare infrastructure crashes using large-scale core dump analysis.
Trust79HypeLow hype23 23Model releaseOrnith-1.0 introduces open-weight models for agentic coding, built on Gemma 4 and Qwen 3.5
Ornith-1.0 is an open-weight model family released under the MIT license by DeepReinforce, the first model release from the organization.
Trust78HypeLow hype24 24Interconnects — Nathan LambertOpen model releases diversify with Zyphra, Cohere, and Poolside additions
Open model ecosystem is diversifying beyond a handful of dominant players, with contributions from niche companies globally.
Trust74HypeLow hype25 25Model releaseOpenAI begins limited preview of GPT-5.6 series with Sol, Terra, and Luna
OpenAI started a limited preview of the GPT‑5.6 series, including flagship Sol, balanced Terra, and low-cost Luna.
Trust79HypeLow hype26 26Google DeepMindGoogle DeepMind integrates computer-use tooling into Gemini 3.5 Flash
Gemini 3.5 Flash now includes built-in computer-use tooling for agentic tasks across platforms.
Trust79HypeLow hype27 27OpenAIOpenAI backs global standards effort for advanced AI through Appia Foundation
OpenAI announced support for a new global standards effort aimed at guiding the development of advanced AI systems.
Trust72HypeLow hype28 28Research blogResearchers identify role confusion as a fundamental challenge in preventing prompt injection
Researchers describe "role confusion" as a core failure mode where models prioritize text style over content, enabling prompt injection attacks.
Trust79HypeLow hype29 29Simon Willison — weblogResearcher ports 0.2B Moebius image inpainting model to run in-browser via WebGPU
A 0.2B-parameter image inpainting model (Moebius) was converted from PyTorch to ONNX and run in a browser using WebGPU.
Trust84HypeLow hype30 30Interconnects — Nathan LambertGLM-5.2 release sparks community praise as a step-change open-weight coding agent
Z.ai released GLM-5.2 on June 13, 2026, to GLM Coding Plan members, with public weights and a blog post on June 16, 2026.
Trust71HypeSome hype31 31MIT Technology Review — AIStartup claims sparse-attention LLM rivals top dense models on coding benchmarks
Subquadratic, a Miami-based AI startup, claims its sparse-attention LLM SubQ matches top coding performance while using far less compute.
Trust71HypeSome hype32 32New model releaseZ.ai releases GLM-5.2, a 753B-parameter open-weights text-only LLM with 1M token context
Chinese lab Z.ai released GLM-5.2 as open weights under an MIT license on June 16, 2026, following a subscriber preview on June 13.
Trust79HypeLow hype33 33arXiv preprintDeepSeek-V4 series introduces two MoE models with million-token context support and efficiency gains
Two new Mixture-of-Experts models, DeepSeek-V4-Pro and DeepSeek-V4-Flash, support one-million-token contexts with improved efficiency.
Trust79HypeLow hype34 34OpenAIOpenAI adds spend controls and usage analytics for ChatGPT Enterprise
OpenAI introduced spend controls and usage analytics for ChatGPT Enterprise to help organizations manage costs.
Trust74HypeLow hype35 35Google DeepMindGoogle DeepMind’s AMIE matches primary care physicians in disease management study
AMIE, Google DeepMind’s research AI system for medical reasoning, matched primary care physicians in overall disease management reasoning in a blinded study with patient actors.
Trust78HypeLow hype36 36Hugging Face Models BlogAllenAI releases MolmoMotion, a language-guided 3D motion forecasting model with new dataset and benchmark
MolmoMotion predicts future 3D trajectories of object points from video frames and language instructions, outperforming prior motion forecasting methods.
Trust79HypeLow hype37 37Hugging FaceGLM-5.2 released with 1M-token context and long-horizon coding improvements
GLM-5.2 is an open-source model optimized for long-horizon coding tasks with a solid 1M-token context.
Trust79HypeLow hype38 38OpenAIOpenAI unveils Deployment Simulation to predict model behavior before release
OpenAI introduced Deployment Simulation, a method to predict AI model behavior before deployment.
Trust72HypeLow hype39 39Audio ModelsStability AI releases Stable Audio 3.0 with models capable of generating six-minute compositions
Stability AI released Stable Audio 3.0, a family of four models ranging from 459M to 2.7B parameters, with top models generating music up to 6 minutes 20 seconds long—double the length of its 2024 predecessor.
Trust52HypeSome hype40 40Remote SensingAllen Institute releases OlmoEarth v1.1, a satellite imagery model that cuts inference costs threefold
Allen Institute released OlmoEarth v1.1, an update to its transformer-based remote sensing model designed for satellite imagery analysis.
Trust74HypeLow hype41 41SearchGoogle's AI Mode search feature surpasses one billion monthly active users one year after U.S. launch
Google reported that AI Mode, its conversational search feature introduced one year ago in the U.S., has reached over one billion monthly active users globally and queries have doubled every quarter since launch.
Trust67HypeSome hype42 42ModelsOpenAI releases GPT-Realtime-2 voice model with expanded reasoning and 128K context window
OpenAI launched GPT-Realtime-2, positioning it as a voice-to-speech model with 'GPT-5-class reasoning' capable of tool use, interruption recovery, and longer conversations via expanded 128K context window.
Trust69HypeSome hype43 43OpenAI Product LaunchOpenAI introduces GPT-5.5-Cyber for verified security researchers
OpenAI announced GPT-5.5 and GPT-5.5-Cyber as part of its Trusted Access for Cyber initiative
Trust74HypeLow hype44 44Reasoning ModelZyphra releases ZAYA1-8B, a 700M-parameter reasoning model competitive with larger alternatives
Zyphra published a technical report on ZAYA1-8B, a reasoning-focused mixture-of-experts model with 700M active parameters and 8B total parameters trained entirely on AMD compute infrastructure.
Trust76HypeLow hype45 45OpenAI APIOpenAI releases realtime voice models with reasoning and translation capabilities
OpenAI has published new realtime voice models for developers to use through its API.
Trust75HypeSome hype46 46SDK ReleaseAnthropic releases Python SDK v0.100.0 with managed agents and webhook support
Anthropic released version 0.100.0 of its Python SDK on May 6, 2026. The release adds support for managed agents, multiagents, outcomes, webhooks, and vault validation. Bug fixes include adjustments to webhook configuration.
Trust79HypeLow hype47 47SDK ReleaseOpenAI Python SDK v2.34.0 adds Admin API key support and auth improvements
OpenAI released v2.34.0 of its official Python SDK on May 4, 2026, adding support for per-endpoint Admin API keys and bearer-free admin authentication.
Trust79HypeLow hype48 48InfrastructureOpenAI advances Stargate compute infrastructure to support artificial intelligence scaling
OpenAI is scaling Stargate, its compute infrastructure initiative, to expand data center capacity
Trust63HypeSome hype49 49Google InfrastructureGoogle outlines TPU capabilities and newest generation specifications
Google describes TPUs as custom chips designed specifically to run AI models, emphasizing their ability to perform complex mathematical operations at scale.
Trust76HypeSome hype50 50AWS IntegrationOpenAI models and Managed Agents now available on AWS
OpenAI announced its GPT models, Codex, and Managed Agents are now available on AWS infrastructure.
Trust74HypeLow hype